Generation of Classification Rules using Artificial Immune System for Fault Diagnosis

This paper presents an artificial immune system based classification rules generation for fault diagnosis of induction motors. To implement the proposed method effectively, a feature extraction and fuzzificiation processes are used for choosing fault-related attributes from motor current signals. The idea behind the method is mainly based on both concepts of data mining and artificial immune system. Association rule set is generated using clonal selection based on confidence and support measures of each rule. Afterwards, an efficiency evaluation method is utilized to construct memory set of classification rules. Each rule is evaluated based on three measures, sensitivity, simplicity, and coverage, to select an optimal rule for classification. The proposed approach was experimentally implemented on a 0.37 kW induction motor and its performance verified on various working conditions of the induction motors. The performance results have shown that a high accuracy rate has been achieved.